You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

EgoManipulation RGB Stereo IMU Raw

Open In Colab

EgoManipulation RGB Stereo IMU Raw is a collection of egocentric manipulation recordings captured using a head-mounted RGB camera, a stereo monochrome camera pair, and an inertial measurement unit (IMU). All sensors share a common hardware device-clock timebase.

  • A Parquet catalog provides recording-level metadata and previews for browsing the dataset without downloading the full MCAP collection.
  • Recordings are stored as MCAP files, with one selected recording chunk per file.
  • Original sensor timestamps are preserved to support offline synchronization, stereo depth estimation, visual-inertial odometry, etc.
  • Camera calibration and sensor extrinsics are included in each recording's session metadata.

Repository layout

README.md
data/
  <task_label>_01.mcap
  <task_label>_02.mcap
catalog/
  recordings.parquet

MCAP filenames use lowercase task labels with spaces replaced by underscores, followed by a two-digit recording number.

Dataset catalog

catalog/recordings.parquet contains one row per MCAP recording. The Hugging Face Dataset Viewer displays this catalog.

Data MCAP format

Each MCAP contains Protobuf messages with the following topics:

Topic Schema Description
/meta/session_meta dataset.SessionMetadata Recording metadata, sensor configuration, camera calibration, and extrinsics
/annotation/task_label annotation.TaskLabel Manipulation task label
/sensor/camera/rgb/image foxglove.RawImage (rgb8) RGB frames (CAM_A)
/sensor/camera/left/image foxglove.RawImage (mono8) Left monochrome stereo frames (CAM_B)
/sensor/camera/right/image foxglove.RawImage (mono8) Right monochrome stereo frames (CAM_C)
/sensor/imu/accelerometer foxglove.Vector3 Accelerometer measurements (IMU, m/s²)
/sensor/imu/gyroscope foxglove.Vector3 Gyroscope measurements (IMU, rad/s)

All topics use Protobuf message encoding, and their Protobuf schemas are embedded directly in each MCAP file. Camera topics include a frame_id channel metadata field identifying CAM_A, CAM_B, or CAM_C. IMU topics use frame_id: IMU and declare measurement units in channel metadata.

Camera images are stored as decoded pixel data: rgb8 for the RGB stream and mono8 for the monochrome stereo streams. MCAP files can be read using a compatible MCAP reader.

Camera calibration and rectification

  • RGB (CAM_A): Frames are already undistorted.
  • Stereo left (CAM_B) and right (CAM_C): Frames retain their original, unrectified and distorted camera geometry. They have not been undistorted or rectified.
  • Calibration metadata: Camera intrinsics, distortion coefficients, and camera-to-camera extrinsics are included in the session metadata.

The left and right stereo streams can be undistorted and rectified as needed using the provided calibration parameters.

Data contract

Published recordings are quality-checked against the following requirements:

Property Requirement
Shared timebase All camera and IMU sensors use timestamps from a common on-device hardware clock.
Camera synchronization Corresponding RGB, stereo-left, and stereo-right frames are synchronized within 1 ms.
IMU boundary alignment Accelerometer and gyroscope streams begin 0–2.5 ms before the first vision frame and end 0–2.5 ms after the last vision frame, ensuring continuous IMU coverage throughout the recording.
Vision frame rate 30 FPS for RGB, stereo-left, and stereo-right streams.
IMU sample rate 400 Hz for accelerometer and gyroscope streams.
Vision drop tolerance At most 1% dropped frames per camera stream.
IMU drop tolerance At most 2% dropped samples per IMU stream.

Timestamps

Original sensor capture timestamps are preserved without resampling or interpolation. The common hardware-clock timebase enables direct temporal alignment across camera and IMU streams.

The 1 ms synchronization tolerance applies to corresponding camera frames. The IMU boundary requirement ensures that both accelerometer and gyroscope streams start 0–2.5 ms before the first vision frame and extend 0–2.5 ms beyond the last vision frame. This guarantees IMU timestamp coverage across the full vision recording interval, but does not imply that individual IMU samples coincide with camera frames.

Downloads last month
335